BACKGROUND OF THE INVENTION
1. Field of the Invention
[0001] The present invention relates to a method for fabricating and assembling integrated
circuit or memory chips on a circuit substrate or printed wiring board and, more particularly,
for stacking such chips into a contacting layered stack, such that their planes lie
parallel to and in contact with the plane of the circuit substrate or printed wiring
board, and to the parallelly stacked chip assembly manufactured by the method.
2. Description of Related Art and Other Considerations
[0002] Conventionally, integrated circuit and memory chips are mounted singly in a hermetic
package, and the package leads are interconnected to a substrate or a printed circuit
board. Alternately, the chips are mounted onto a circuit board in a hermetic hybrid
package whose external leads are interconnected to the printed wiring board.
[0003] When space and volume are at a premium, use of such conventional mounting arrangements
results in a less than optimal overall circuit density and a lower circuit speed.
Specifically, presently used mounting arrangements consume relatively large amounts
of space and volume, and thus lessen circuit density. Further, the relatively large
spacing between the chips lowers circuit speed and increases capacitances, thereby
causing slower signal propagation through the circuit.
[0004] These problems have been recognized in the packaging technology, as exemplified by
United States patents 4,525,921 and 4,764,846. There, chips are stacked at right angles
to the substrate and interconnected with edge metalizing circuitry. On one edge or
face of the stack, the circuit lines terminate at metal bumps which are used to interconnect
the stack directly to a substrate and its corresponding pattern of circuit pads. The
chips are thus oriented so that their planes lie normal to the plane of the substrate.
[0005] The aim of these patents is to achieve a high density in the stack. However, as disclosed,
for example, in patent 4,764,846, to enable the mounting of its individual chips and
their carriers to a stack-carrying substrate, the leads from the chips must be extended
to a single side, which consumes a relatively large area, thereby militating against
the desired aim of high density.
[0006] It is also possible further to increase density by lapping material from the back
side of the wafer prior to its being diced into thinned chips. This thinning is expensive,
and adds to the cost of the stack.
[0007] The bump mounting method of the stack is also costly, and its use in producing reliable
connections is uncertain for use in such environments as airborne systems.
SUMMARY OF THE INVENTION
[0008] The present invention overcomes these and other problems and provides for a still
higher density over such above described conventional methods by stacking two or more
integrated circuit or memory chips on a circuit substrate or printed wiring board
in such a manner that the planes of the chips are parallelly, rather than normally,
oriented to the substrate or wiring board. The chips can be interconnected as desired,
not only from one edge, but also about their entire periphery and thence, preferably
by ribbon bonds, to the substrate or wiring hoard. The thus assembled arrangement
is hermetically sealed by coatings of passivation and encapsulant.
[0009] Such chips are oversized, as distinguished from chips conventionally diced from wafers.
Specifically, each chip is larger than an individual wafer circuit, that is, each
wafer portion which is selected to be formed into a chip has a size that is larger
than the individual wafer circuit which it incorporates, thus overlapping adjacent
circuits.
[0010] Several advantages are derived from this arrangement. As compared to conventional
designs, overall circuit density is further increased by holding the space and volume
used to a minimum, by a factor, for example, of about five times over that disclosed
in above discussed patent 4,764,846. Because the spacing between the chips is minimized
circuit speed is increased and line impedance and capacitance is correspondingly lowered,
thereby permitting higher signal propagation rates through the circuit. The cost of
fabrication is lowered and the use of bump mounting is avoided, thus resulting in
reliable interconnections. Thermal dissipation is improved. Special hermetic packaging
may be avoided, if desired.
[0011] Other aims and advantages, as well as a more complete understanding of the present
invention, will appear from the following explanation of exemplary embodiments and
the accompanying drawings thereof.
BRIEF DESCRIPTION OF THE DRAWINGS
[0012]
FIG. 1 is a flow diagram of a typical fabrication and assembly procedure for producing
products encompassed by the present invention;
FIG. 2 is a representation of a typical full wafer which is modified to show a commencing
step in the fabrication and assembly procedure of the present invention, using good
circuits, designated "G", for incorporation into each chip whose size is greater than
the good circuit which it incorporates;
FIG. 3 illustrates one embodiment of the present invention comprising a stacked chip
assembly before being mounted on a circuit substrate; and
FIGS. 4-7 depict other embodiments of the present invention, showing alternate construction
methods which have been devised to meet special requirements for mounting memory and
other chips on a circuit substrate or a printed wiring board.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0013] Reference is first made to FIGS. 1 and 2, the former defining a flow diagram of the
fabrication and assembly procedure of the present invention, and specifically to the
steps designated by indicia 12 and 14. There, indicium 12 represents the method step
where a set of dielectric masks is designed to define electrical connections based
upon a desired electronic function or functions, and indicium 14 indicates the step
in which a wafer 16 (see also FIG. 2) is tested by conventional methods to identify
the locations on the wafer of good circuits 18 (also designated in FIG. 2 by indicium
"G") and bad circuits 20. It is typical to dot the wafer where bad circuits are found,
rather than to mark the good circuits with a "G" designation.
[0014] In conformance with the present invention, a dicing scheme is defined for wafer 16
(step 22, FIG. 1), and defines locations 24 of oversized chips 25 which are to be
diced from the wafer, as shown in FIG. 2. Dicing occurs preferably in two steps. First,
the wafer is sawed, e.g., horizontally as shown by dashed lines 27, in such a manner
as to avoid slicing through any oversized chip 25, and thus to separate the wafer
into rows of chips 25, such as illustrated in FIG. 2. Then, each row is further sawed
at a normal, i.e., vertically to the previously made horizontal saw lines 27, to form
individual oversized chips 25.
[0015] In the dicing scheme of step 22 (FIG. 1), wherever possible to maximize use of the
wafer material, locations 24 are selected to overlie bad circuits but, if necessary
or unavoidable, over good circuits. As an important step in the present invention
(step 26 of FIG. 1), the good circuits are masked to protect their input/output pads
and wafer 16 is passivated with an impervious, inorganic coating to provide the necessary
hermeticity of the chip circuitry. After further appropriate masking, dielectric material
is applied to the wafer (step 28 of FIG. 1). This dielectric material may comprise
a thin overcoat of organic resin. A pattern of input/output relocation metal is then
applied to the wafer to define leads 32, and thus to extend input/output pads 34 of
each chip shown in FIG. 2. The performance of this step is noted in step 30 of FIG.
1.
[0016] At this point, the processing of the wafer varies, depending upon whether the chips,
which are to be diced from the wafer comprise lower or topmost chips. This variation
in processing, as denoted in FIG. 1, is represented as diverging processing steps
for lower chips 36 and topmost chips 38. Those chips 38, which are intended to function
as top chips, require two additional processing steps 40 and 42. As depicted in step
40, ribbon bondable metal is applied to peripheral pads 34 (FIG. 2) and, as instructed
in step 42, these pads are masked. If desired, wire leads may be used in place of
ribbon leads, although the latter is preferred. As required by respective steps 46
and 48 for the lower and the top chips, the wafer is passivated, followed by a dicing
of the wafer into lower and topmost chips.
[0017] As set forth in step 52, for each electric device, the lower chips and the topmost
chip are stacked and bonded together into one of the assemblies illustrated in FIGS.
3-7.
[0018] For only the embodiment disclosed in FIG. 3, the three steps identified in FIG. 1
by indicia 54, 56 and 58 and bracketed by indicium 60 are practiced to produce an
assembly 62. In step 58, for example, as it is understood in the art, the term "clear"
refers to the step of removing passivation from interconnection pads, if the entire
wafer has been passivated in step 46 prior to its being diced, and the term "metalize"
refers to the application of suitable metallurgy to facilitate ribbon bonding, if
necessary.
[0019] Here, assembly 62 comprises a top chip 64 and lower chips 66 of equal planar dimensions.
Top chip 64 comprises one of chips 25 shown in FIG. 2 which has been processed according
to steps 38, 40, 42, 46 and 50 of FIG. 1 so that it includes leads 32' and terminal
pads 34'. Lower chips 66, like top chip 64, comprise several of chips 25 of FIG. 2
which have been processed according to steps 36, 46 and 48 of FIG. 1 so that they
include leads (not shown, but like leads 32 terminating in pads 34).
[0020] All of chips 64 and 66, after having been stacked and bonded together as subassembly
62, are ground at their respective edges 68 and on all sides on which connections
are to be made, in order to ensure that all pads (e.g., pads 34' of topmost chip 64
and like pads on lower chips 66) are exposed and made flush in the planes of their
respective sides, in accordance with step 54 of FIG. 1. Because it is desirable to
increase the amount of metal of the exposed pads on lower chips 66, metalization 69
is added to these exposed pads. Interconnects 70 are then formed by conventional metalization
techniques between or among selected pads (e.g., pads 34') and metalization 69 and,
further, top pads 34' are metalized or cleared, as required by the electrical function,
pursuant to steps 56 and 58.
[0021] Stacked and bonded subassembly 62 is then mounted on and affixed to a base 72, having
pads 74, as required by step 76, and ribbon leads 78 are bonded respectively to pads
34' and to pads 74, as represented by step 80, to form an assembly 82.
[0022] Assembly 82 is then electrically tested according to step 84 to determine if its
electrical function operates as intended. Should the test prove successful, assembly
82 is then mounted on and ribbon bonded to a fineline or other appropriate circuit
board in accordance with steps 86 and 88. If the test is not successful, assembly
82 is returned for further processing or discarded, as appropriate. Successfully tested
constructions are then encapsulated, if needed, as noted by step 90.
[0023] Assemblies 92, 94, 96 and 98, respectively shown in FIGS. 4-7, are also processed
in accordance with the steps depicted in FIG. 1, except for steps 54, 56 and 58 thereof.
[0024] In the method for fabricating assembly 92 of FIG. 4, a substrate circuit 100 is provided
with sufficient pads 102 to permit direct connection to the circuit of individual
chips 104 which are stacked upon one another. No connections are bade between the
chips, but the chips can use selected pads in common on the substrate circuit. This
technique permits a more conventional assembly to be made using existing processes,
and does not require the development of new technologies or interconnecting chips
on their vertical edges as shown in FIG. 3.
[0025] As disclosed with respect to assembly 94 of FIG. 5, in those cases where it is desirable
to mount a chip 106 in very close proximity to another chip 108, one expedient method
is to mount chip 106 atop chip 108, and to interconnect upper chip 106 directly to
lower chip 108 without the necessity of interconnections being made to a substrate
as an intermediary circuit path. To accomplish this, it is necessary that upper chip
106 be smaller than lower chip 108, so as to provide room for pads on the lower chip
to accommodate interconnecting wires or ribbons 110 from the upper chip. Assembly
94 is then mounted and electrically coupled to a substrate (not shown) by groups of
ribbon leads 112.
[0026] As illustrated in FIG. 6, it is possible also to mount several chips in a stack 114,
like chips 104 shown in FIG. 4, where a group of upper chips 116 are smaller than
chip 118 below, to permit accommodation of interconnecting pads on the lowest chip
for all of the uppermost chips. The lowest chip has pads at its periphery to permit
its connection by ribbon leads 120 to a substrate (not shown).
[0027] Yet another method is shown in FIG. 7 where an upper chip 122 is flip-chip mounted
with bump pads 124 to a lower chip 126. Here, the matched coefficient of thermal expansion
of silicon on silicon can be used to advantage.
[0028] Although the invention has been described with respect to particular embodiments
thereof, it should be realized that various changes and modifications may be made
therein without departing from the spirit and scope of the invention.
1. A method for manufacture of a stacked chip assembly by fabricating, assembling and
interconnecting at least two integrated circuit/memory chips in a stacked upper and
lower chip arrangement to a circuit substrate/printed wiring board support, comprising
the steps of:
testing a full wafer for determining operable and inoperable circuits therein;
providing a dicing scheme for defining chips whose individual sizes are respectively
larger than individual ones of the circuits and for positioning individual ones of
the chips on selected ones of the operable circuits, thus overlapping the chips onto
ones of the circuits adjacent the selected circuits, in preparation for dicing the
wafer into the upper and lower chips;
applying a dielectric layer mask to the wafer in accordance with circuit designs
defining the constituent parts of the stacked chip assembly;
applying input/output relocation metal to portions of the wafer exposed through
the mask;
for the top chip, applying ribbon bondable metal to peripheral pads on the wafer,
and masking the pads;
passivating the wafer;
dicing the wafer into the upper and lower chips;
assembling and bonding the chips, one atop and in contact with an adjacent chip,
into a layered stack of chips;
mounting the stack on the base;
ribbon bonding the stack to a base;
electrically testing the stacked and bonded chips;
mounting the stacked and bonded chips and base to the support;
ribbon bonding the stacked and bonded chips and base to the support; and
encapsulating the stacked and bonded chips, base and support assembly.
2. A method according to claim 1 further including the steps, intermediate said steps
of assembling and bonding the chips into the stack and of mounting the stack on the
base, of:
grinding edges of the chips to expose electrical contact points;
metallizing the edges at selected ones of the contact points;
metallizing selected ones of the pads on the uppermost chip; and
performing said stack to base ribbon bonding step by bonding ribbon leads from
the metallized edges and pads to designated contact areas on the base.
3. A method for fabricating an assembly of chips having large surfaces relative to their
thicknesses, comprising the steps of:
stacking at least two integrated circuit chips one atop the other, in such a manner
that the large surface of one chip is in contact with the large surface of its adjacent
chip;
assembling the stacked chips on a surface of a base such that the large surface
of one of the chips is in contact with the surface of the base; and
electrically interconnecting the chips to the base.
4. A method according to claim 3 further comprising the prestacking step of forming each
of the chips from a wafer having a plurality of circuits thereon by selecting wafer
portions defined by individual chip sizes which are respectively larger than individual
ones of the circuits and by positioning individual ones of the chips on selected operable
ones of the circuits, thus overlapping the chips onto ones of the circuits adjacent
the selected circuits.
5. A method according to claim 4 further comprising the step of positioning contacts
fully about the periphery of each of the chips.
6. A method according to claim 4 in which the stacked chips have equal dimensions, further
comprising the steps of electrically interconnecting the chips at their edges and
electrically interconnecting the uppermost chip and the base.
7. A method according to claim 4 in which the stacked chips have equal dimensions, further
comprising the step of electrically interconnecting the chips to the base.
8. A method according to claim 4 in which the stacked chips have unequal dimensions,
further comprising the steps of stacking at least one of those chips having lesser
dimensions atop at least another of those chips having relatively larger dimensions,
and electrically interconnecting the chips successively to the lower chip and therefrom
to the base.
9. A method for fabricating an assembly of chips comprising the steps of:
stacking at least two integrated circuit chips one atop the other;
assembling the stacked chips parallelly on a base; and
electrically interconnecting the chips to the base.
10. A method according to claim 9 further comprising the prestacking step of forming each
of the chips from a wafer having a plurality of circuits thereon by selecting wafer
portions defined by individual chip sizes which are respectively larger than individual
ones of the circuits and by positioning individual ones of the chips on selected operable
ones of the circuits, thus overlapping the chips onto ones of the circuits adjacent
the selected circuits.
11. A method according to claim 9 further comprising the step of positioning contacts
fully about the periphery of each of the chips.
12. A method according to claim 10 further comprising the step of forming interconnections
among the chips prior to said chip to base interconnecting step.
13. A method according to claim 12 wherein said chip to base interconnecting step comprises
the step of forming connections from an uppermost one of the chips to the base.
14. An assembly of chips comprising at least two integrated circuit chips stacked one
atop the other, assembled parallelly on a base, and electrically interconnected to
said base.
15. An assembly according to claim 14 further comprising means for forming each of said
chips from a wafer having a plurality of circuits thereon in which wafer portions
are selected by individual chip sizes which are respectively larger than individual
ones of said circuits and in which individual ones of said chips are positioned on
selected operable ones of said circuits, thus overlapping said chips onto ones of
said circuits adjacent said selected circuits.
16. An assembly according to claim 14 further comprising contacts positioned fully about
the periphery of each of said chips.
17. An assembly according to claim 15 in which each of said chips are provided with edge
electrical contacts extending fully about its periphery, further comprising electrical
interconnections among and between said edge contacts of each of said chips for electrically
interconnecting said chips together.
18. An assembly according to claim 17 further comprising electrical connections extending
from an uppermost one of said chips to said base.
19. An assembly according to claim 17 in which said stacked chips have equal dimensions,
said electrical interconnections among said chips are positioned at the edges of said
chips, and electrical interconnections between the uppermost chip and said base.
20. An assembly according to claim 17 in which said stacked chips have unequal dimensions,
with at least one of those chips having lesser dimensions being stacked atop at least
another of those chips having relatively larger dimensions, and further comprising
electrical interconnections between successive ones of said upper chips to said lower
chips and thence to said base.